| 摘要 |
The apparatus first generates an integer motion vector MVO with absolute error AEO based on a comparison of pixel blocks of the current image(fig 1B,1) with the pixels of the previous frame (2) in the form of a first search window (fig 1A,SW1'). A second search window (SW2') is formed based on the integer pixel motion vector MVO and an interpolator 100 produces half pixels b 0 1 ,b 1 0 ,b. Motion estimation is carried out between these half pixels and the reference pixels of the previous frame a(x,y),a'(x,y). Absolute errors are generated by calculating the differences between the pixels of the previous frame and the half pixels 221-228,(fig 6,251-253) and passing them to the absolute value calculator (253) before being stored in accumulators 231-238. The comparing and selecting means then compares the accumulated errors AE1-AE8 to the original error AEO and outputs a motion vector corresponding to the minimum absolute error(see table 1,page10). |
